I see this question pop up a lot, so figured I'd break down how I think about it after running both ZHA and Z2M on different setups. Zigbee itself Zigbee isn't software you choose, it's the wireless protocol your devices speak. Every bulb, sensor, and plug that says "Zigbee" on the box is talking the same low-power mesh protocol regardless of what's running on your Home Assistant server. The protocol handles the radio layer, pairing, and mesh routing between devices. What you do choose is the software that sits between your Zigbee coordinator (the USB stick or gateway) and Home Assistant, translating Zigbee traffic into something HA understands. That's where ZHA and Z2M come in. ZHA (Zigbee Home Automation) ZHA is Home Assistant's built-in integration. No separate app, no extra container, no MQTT broker to manage. You plug in a coordinator, add the integration, and pair devices straight from the HA UI. What I like about it: Setup is genuinely the fastest of the two. Fewer moving parts means fewer things to misconfigure. Device support is handled through quirks, community-contributed device profiles that patch weird or non-standard implementations. Everything lives inside HA, so there's no second web UI to check when something's not reporting. Where it falls short: Network visibility is limited. There's no built-in map showing which devices are routing through which repeaters, so diagnosing mesh problems is more guesswork. Some devices get supported later than they do on Z2M, since the quirks system depends on community contributions catching up. Zigbee2MQTT (Z2M) Z2M runs as a separate add-on (or standalone service) and talks to HA over MQTT rather than being built into HA directly. What I like about it: The network map in the Z2M frontend is genuinely useful. You can see LQI (link quality) between every device and repeater, which makes tracking down a flaky sensor way faster than guessing. Broader and often faster device support, especially for newer or less common devices. More granular control over channel settings, reporting intervals, and per-device configuration. Where it falls short: You're now running MQTT as a dependency, which is one more thing that can go wrong or need troubleshooting. Slightly more setup overhead upfront, especially if you've never touched an MQTT broker before. So which one do you pick? If you want the simplest possible setup and don't care about deep network diagnostics, ZHA is the easier starting point. If you've got a larger mesh, want to actually see what's happening between your coordinator and your devices, or you're chasing down flaky connections, Z2M's visibility is worth the extra setup step. I run Z2M mainly because I lean on the network map constantly when troubleshooting range issues, but I know plenty of people happily running dozens of devices on ZHA without ever needing that level of detail. Neither one is "wrong," they're just optimized for different priorities. Curious what others are running and why you landed where you did.

One of the most common frustrations in the Home Assistant community right now: Zigbee devices that pair fine, work for a day or two, then drop off and show as unavailable, sometimes coming back on their own, sometimes not. The generic answer you'll find everywhere is "improve your mesh." That's not wrong, but it's not useful on its own. Here's what's actually worth checking, in order: Coordinator placement first, mesh second The coordinator is the root of the entire Zigbee mesh. If it's plugged directly into a USB port on your Home Assistant machine, USB 3.0 interference is a likely culprit. It broadcasts in the 2.4GHz range and drowns out Zigbee signals. Fix: use a USB 2.0 extension cable (1–2 metres) to move the coordinator away from the machine. This alone solves the problem for a large number of people. Powered devices vs. battery devices Battery-powered devices (sensors, buttons) are end devices — they don't route traffic. Only mains-powered devices (plugs, bulbs) act as routers and extend your mesh. If most of your devices are battery-powered, your mesh has almost no routing capacity. Add at least 2–3 mains-powered devices within range of the coordinator before assuming anything else is broken. Channel conflicts with Wi-Fi Zigbee and 2.4GHz Wi-Fi share spectrum. Zigbee channels 11–13 overlap heavily with Wi-Fi channel 1; Zigbee channels 25–26 overlap with Wi-Fi channel 11. If you haven't set your Zigbee channel deliberately, check what your Wi-Fi is using and pick a Zigbee channel that doesn't overlap. Zigbee channel 15, 20, or 25 are usually the safest depending on your Wi-Fi setup. ZHA vs. Zigbee2MQTT coordinator handling Both ZHA and Zigbee2MQTT handle coordinator recovery differently after a Home Assistant restart or power cut. If devices drop specifically after HA restarts, check whether your coordinator firmware is current — outdated firmware on Sonoff Zigbee 3.0 USB dongles in particular is a known cause of post-restart drops. Check LQI before assuming range In Zigbee2MQTT, go to the map view and look at the LQI (Link Quality Indicator) values on the connections to your dropping devices. Anything below 50 is weak; below 20 is effectively broken. This tells you whether it's a range/routing problem or something else entirely. What coordinator are you running, and are the drops happening on specific device types or randomly across the mesh? Drop your setup below, happy to help diagnose.

I've been going through everything Apple announced for the Home app at WWDC, and the camera changes are the most interesting thing HomeKit has done in years. But I want to argue about the part nobody's really pinning down: does this actually match the paid cloud tiers, or just look like it does? What's actually coming Confirmed for the fall releases (iOS 27 / tvOS 27 / HomePod 27): AI-written summaries — instead of "motion detected," you get "a person walked up to the door" or "a dog was in the yard." Toggleable per camera. Cross-camera event grouping driveway cam + doorbell cam get stitched into one event with one summary, automatically. No manual linking. Rolling notifications — related alerts collapse into a single notification that updates in real time instead of blowing up your lock screen. Searchable footage — type "package delivery" and it jumps to the clip instead of you scrubbing a 3-hour timeline. Long-press previews + 4K recording (4K doesn't even need Apple Intelligence). On paper, that's feature parity with Nest Aware and Ring Protect's "smart" tiers. Here's where I want to push back on the hype Two things I keep seeing repeated that aren't quite right: "It's all local, no cloud." Not exactly. The AI runs through your home hub (Apple TV or HomePod), not purely on your phone and HomeKit Secure Video still needs an iCloud+ subscription. So it's "on your hardware" but it is not subscription-free. You're still paying Apple monthly; the money just goes to iCloud instead of Ring. Nobody's tested the accuracy yet. These features ship in the fall — everything floating around right now is announcement footage, not real-world review. Google shipped near-identical AI summaries last year and raised prices, and reviewers still griped about wrong summaries. Apple gets no free pass here until people actually run it. The actual debate So the real question isn't "are these cool" (they are). It's: If HomeKit Secure Video now does named people, package detection, multi-cam event stitching, and natural-language search all bundled into an iCloud+ plan you probably already pay for what's left justifying a separate Ring or Nest subscription? Counter-argument I keep landing on: the iCloud+ camera limits are stingy (50GB = one camera, 200GB = five), and Nest/Ring still win on hardware range and battery cameras. Where do you land? Specifically: Anyone in the beta actually seeing the AI summaries fire yet? How accurate? Is the iCloud+ per-tier camera cap a dealbreaker vs. a flat Nest Aware plan? Does "processed on my own Apple TV" actually matter to you privacy-wise, or is that overrated when the footage is in iCloud anyway?